EP2943935A1 - Estimation de mouvement d'une image - Google Patents
Estimation de mouvement d'une imageInfo
- Publication number
- EP2943935A1 EP2943935A1 EP14703136.3A EP14703136A EP2943935A1 EP 2943935 A1 EP2943935 A1 EP 2943935A1 EP 14703136 A EP14703136 A EP 14703136A EP 2943935 A1 EP2943935 A1 EP 2943935A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- term
- image
- points
- data
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000033001 locomotion Effects 0.000 title claims abstract description 74
- 239000013598 vector Substances 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims abstract description 67
- 230000033228 biological regulation Effects 0.000 claims description 20
- 230000001788 irregular Effects 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 4
- 230000001052 transient effect Effects 0.000 claims 1
- 238000006073 displacement reaction Methods 0.000 description 24
- 238000004364 calculation method Methods 0.000 description 17
- 239000011159 matrix material Substances 0.000 description 17
- 238000001914 filtration Methods 0.000 description 5
- 230000002123 temporal effect Effects 0.000 description 4
- 230000003750 conditioning effect Effects 0.000 description 3
- 238000009499 grossing Methods 0.000 description 3
- 230000001364 causal effect Effects 0.000 description 2
- 230000003247 decreasing effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 150000003839 salts Chemical class 0.000 description 2
- LZDYZEGISBDSDP-UHFFFAOYSA-N 2-(1-ethylaziridin-1-ium-1-yl)ethanol Chemical compound OCC[N+]1(CC)CC1 LZDYZEGISBDSDP-UHFFFAOYSA-N 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/513—Processing of motion vectors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/269—Analysis of motion using gradient-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/254—Analysis of motion involving subtraction of images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/277—Analysis of motion involving stochastic approaches, e.g. using Kalman filters
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
- H04N19/43—Hardware specially adapted for motion estimation or compensation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N5/00—Details of television systems
- H04N5/14—Picture signal circuitry for video frequency region
- H04N5/144—Movement detection
- H04N5/145—Movement estimation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
Definitions
- the invention relates to the field of image processing, and more particularly to the field of motion estimation within a captured image sequence.
- a sensor When a sensor captures a sequence of successive images, for example in the case of a video, it is known to perform an overall motion estimation between images. This motion estimation is intended to determine the overall motion affecting the sequence of images between successive pairs of images of the sequence and may correspond to the determination of the movement of the line of sight of the sensor used.
- Such estimate global motion notably allows the implementation of an image stabilization, image noise suppression, or algorithms super ⁇ resolution.
- this type of processing can be substantially disturbed when the captured scene includes objects that are movable in the scene during the sequence of images in question.
- Image processing systems are then based on the implementation of a motion estimation no longer global but dense, also called "local motion estimation”.
- a dense motion estimation consists in estimating the motion at each point P ( x > y, or pixel, of the images of the captured sequence, more precisely, a displacement field d ⁇ x, y) is defined from two images and corresponds to the vectors to be applied respectively to each point of the first image to obtain the corresponding point of the second image.
- This velocity field is also called optical flow and is formed by the set of velocity vectors at a given moment of each point of the image. This velocity field is a function of the derivative of the displacement field with respect to a fixed reference.
- I (x, y, t) I (x + dx, y + dy, t + dt) (_)
- H x >y> t) is a luminous intensity value (hereinafter referred to as "intensity") of a point P (x, y) of the scene at a capture instant t of an image called the preceding image - ⁇ .
- a next image said image is captured at a time t + dt e dt corresponds to the time interval between the two - ⁇ photographs.
- (dx, dy) d (x, y) denotes the displacement vector of the point P ( x > y) between instants ⁇ and t + dt.
- I (x, y, t) I (x, y, t) + [- dx + -dy + - dt] (2)
- Equation (2) can be written as follows:
- the Horn-Schunck method imposes a hypothesis of regularity of the velocity field which limits its efficiency in the presence of discontinuities of the velocity field. Its resolution is also done by gradient descent and is therefore slow.
- Lucas-Kanade method assumes the local constancy of the velocity field. It is robust to noise but has the disadvantage of being sparse, that is to say that it does not make it possible to estimate the velocity fields at the points in the vicinity of which the intensity vector spatial gradient is locally uniform.
- substantially uniform is meant that the intensity vector spatial gradients at the different points of the neighborhood considered are either substantially parallel to each other or substantially zero.
- substantially zero is meant that the norm of a vector spatial gradient of intensity considered is lower than the noise of the image.
- vector intensity intensity gradient and “vector intensity spatial gradient” will also be used interchangeably.
- a first aspect of the invention proposes a motion estimation method in a series of images captured by an image sensor, said series of images comprising at least one preceding image and a next image. , the images having points with which intensity data are associated.
- the method comprises an estimated motion vector estimation step V, performed in a plurality of points of an image chosen from the list constituted by the previous image - ⁇ and the following image.
- the estimation step includes, for each of the points P of said plurality of points of an image, an operation of minimizing a functional J of motion vectors V.
- the functional J comprises a sum of at least one term term term data and a term term regulation term.
- the term of the data is a function of spatio-temporal gradients of intensity data associated with a plurality of points of a neighborhood of a point P.
- the term of the data is a function of the vector spatio-temporal gradients associated intensity data (i.e., each of their scalar components).
- the control term has a finite value at least in each of the points P of said plurality of points of an image at which the minimization of the data term provides an irregular solution.
- a solution is irregular if the minimization of the term of attachment to the data does not make it possible to estimate a single solution.
- the term data includes a product of a term called weighting term and a term term term of attachment to the data,
- the weighting term being a function of a degree of regularity of estimated motion vectors
- the term of attachment to the data being a function of spatio-temporal gradients of intensity data associated with a plurality of points of a neighborhood of a point P.
- the term of attachment to the data is a function of the vector spatio-temporal gradients of associated intensity data (that is to say of each of their scalar components).
- Previous image and “next image” of a series of captured images, two images that follow each other chronologically in the series of images considered. These two images that follow can be consecutive (that is to say, not be separated by any intermediate captured image) or else spaced apart by one or more intermediate captured images. No limitation is attached to the spacing between the previous image and the next image considered here.
- estimate of estimated motion vectors is intended to mean an estimation method making it possible to provide estimated motion vectors for each part of the image, with more or less precision. An image movement can thus be estimated at each point of the image for example.
- This type of motion estimation makes it possible to represent in particular the movement of moving objects that could pass through the images of the series of captured images.
- those skilled in the art are particularly familiar with the Lucas-Kanade and Horn-Schunck methods.
- a first method consists in solving the equation of the least-squares motion on a local window.
- a second method extracted from the article by Berthold KP Horn and Brian G. Schunck, "Determining optical flow” (ARTIFICAL INTELLIGENCE, 1981), consists of a minimization of a functional including a term of attachment to the data comprising the square of the apparent motion equation, and a regularization term including the square of the local variation of the field.
- data refers to the set of intensity data of the points of an image.
- the "end of data” and "term attachment to the given”) are terms based on spatial gradients ⁇ temporal intensity at points located in a neighborhood of a point ( ⁇ ⁇ ').
- the extent of this neighborhood for example the maximum distance of the points included in this neighborhood with respect to the point, as well as a weighting of each of these points in a calculation of term of attachment to the data can be variable according to the conditions of use of the process.
- regulation term is meant a term having a finite value at least in each of the points of the image at which the minimization of the term of the data provides an irregular solution.
- the minimization of the data term provides an uneven solution in homogeneous regions and regions of low angular diversity.
- the homogeneous regions are regions of substantially constant intensity and therefore in which the modules of the vector spatial gradients of the intensities are substantially zero (below a determined threshold, which may correspond to a noise level).
- the associated vector spatial gradients of intensity at different points in the region are substantially parallel.
- the angular diversity is measured for example by the variance of the set of angles defined respectively by a pair of gradients, for all possible pairs of gradients chosen in all of the gradients considered.
- a region of low angular diversity is a region in which the angular diversity of the vector intensity spatial gradients associated with the various points of the region is substantially zero (below a determined threshold, which may correspond to a noise level).
- a finite value of a term is a real or imaginary value, unique and not infinite. Such a finite value may be defined prior to an estimation step or computable by means of a regular function defined over an interval comprising a point.
- a regulation term thus makes it possible to smooth a motion vector field by providing a solution to the problem of minimizing a functional at the levels of the points of an image where the minimization of a term of attachment to the datum makes it impossible to estimate a single solution.
- weighting term is meant a term for weighting the relative importance of terms of attachment to the data and of regulation terms, the weighting being performed as a function of a spatial gradient of the intensity data in a neighborhood of a point .
- the weighting is performed according to a vector spatial gradient of the intensity data in a neighborhood of a point.
- estimated degree of motion vector regularity is meant a function of the variance of the estimated motion vector field.
- the regulation term includes a function of estimated motion vectors V obtained at the preceding iteration n ;
- the weighting term comprises a function of a local variance of estimated motion vectors V obtained at the preceding iteration n ;
- the operation of minimizing the functional comprises least-squares minimization;
- the method further comprises a step of scaling images of the series of images, said scaling step being performed prior to the estimation step of estimated motion vectors V;
- the scaling step and the estimating step are repeated for at least two different scales of images in the series of images;
- the method further comprises a step of resetting the next image, carried out after the scaling step and before the minimization step, the resetting step comprising, for an iteration n + 1 of said steps scaling, resetting and estimating, resetting the next image based on the estimated motion vectors obtained at the previous iteration n ;
- the estimated motion vectors V obtained at an iteration n + 1 are conserved if they satisfy a criterion of estimation reliability and replaced by a function of the estimated motion vectors obtained at the previous iteration n if they do not verify not the said criterion.
- the subject of the invention is also a system for processing a series of images captured by an image sensor, said series of images comprising at least one preceding image - ⁇ and a subsequent image, the images comprising each of the points associated with intensity data,
- the system comprising an estimation circuit adapted to estimate estimated motion vectors V at a plurality of points of an image chosen from the list constituted by the preceding image - ⁇ and the following image ,
- said estimate comprising, for each of the points P of said plurality of points of an image, an operation for minimizing a functional J of motion vectors,
- the functional J comprising a sum of at least one term of the data and a regulation term
- the term of the data being a function of spatio-temporal gradients of intensity data associated with a plurality of points of a neighborhood of a point P,
- the regulation term having at least one finite value at each of the points P of said plurality of points of an image at which the minimization of the term of the data provides an irregular solution
- the term data comprising a product of a weighting term and a term of attachment to the datum, the weighting term being a function of a degree of regularity of estimated motion vectors,
- the term of attachment to the data being a function of spatio-temporal gradients of intensity data associated with a plurality of points of a neighborhood of a point P.
- the subject of the invention is also a computer program comprising instructions capable of implementing a method as described above, during execution of this program by a circuit of a series processing system. images captured by an image sensor.
- the invention finally relates to a non-transitory recording medium on which is stored a computer program as described above.
- FIG. 1 illustrates the main steps of a motion estimation method according to an embodiment of the present invention
- FIG. 2 illustrates an implementation of an estimation method according to an embodiment of the present invention
- FIG. 3 illustrates another implementation of an estimation method according to an embodiment of the present invention
- FIG. 4 illustrates a processing device comprising means adapted to implement a method according to an embodiment of the present invention
- FIGS. 5a, 5b, 5c and 5d respectively illustrate a first source image, a second source image, an estimate of the horizontal velocity field obtained following the execution of a method according to an embodiment of the present invention and a estimation of the vertical velocity field obtained following the execution of a method according to an embodiment of the present invention.
- H (v x , v y ) ⁇ W (xx 0 , y- y Q ) - (x, y) v x (x 0 , y 0 ) + - (x, y) v ( x 0 , y 0 ) + -
- the window function W makes it possible to weight on all the points of the neighborhood ⁇ of the point . It can for example be such that
- Solving the system of least squares equation (5) requires inverting a 2x2 size matrix. This inversion is only possible at points having an angular diversity of spatial gradients of sufficiently high intensity, as specified above, for example points in the vicinity of which there are at least two points including spatial gradients vector of light intensity have substantially distinct directions.
- an image 1 can for example be formed of a clear homogeneous region 2 as well as of a dark homogeneous region 3 of generally rectangular shape.
- a first neighborhood 4 having for example the general shape of a disc centered on one of the vertices of the dark homogeneous region 3, will therefore comprise vector spatial gradients of intensity in two different directions 4a and 4b, for example following the horizontal direction and the vertical direction. A displacement of the dark homogeneous region 3 can then be completely determined in this vicinity.
- a second neighborhood 5, similar in shape to the first neighborhood 4 and located for example at one of the vertical edges of the dark homogeneous 3, comprises a vector spatial gradient of the intensity only in a direction 5a, for example the horizontal direction.
- the displacement of the homogeneous dark region 3 can not be completely determined.
- a third neighborhood 6 also of similar shape to the first neighborhood 4, is for example located entirely in the light homogeneous region 2. In this vicinity 6, the vector spatial gradient of the intensity is substantially zero and the estimate of the field of displacement is therefore impossible.
- This functional has three terms: a term of attachment to the data, a term of weighting according to the degree of regularity of the estimated field and a term of regulation, which will now be detailed.
- the functional consists of the product of the weighting term and the term of attachment to the datum, to which is added the term of regulation.
- the preceding equation (6) is a term of attachment to the datum which takes into account the local constancy of the velocity field in a neighborhood of the point (o '), similar to the functional H of the equation (5). ) above.
- the minimization of this term makes it possible to estimate a velocity field as a function of the spatio-temporal gradients of the luminous intensity.
- ⁇ ( v x ' v > -) of the preceding equation (6) is a weighting term as a function of the degree of regularity of the estimated field of the term of attachment to the datum. This term makes it possible to reduce the importance of the term of attachment to the data in the homogeneous and one-dimensional regions.
- a function of the form ⁇ p (var (v x , v y )) can be used for the term ⁇ ( V * ' V J, the function ⁇ being a decreasing function with a value between 0 and 1 and var ( v * > v y ) being the variance of the speed field at the point or in a neighborhood of the point.
- the variance var ( v x ' v > -) is high and the weighting term is close to 0. If, on the contrary, the estimate is regular, the variance is low and the weighting term will be close to 1.
- a weighting term can be used, where v x and v are velocity field averages over a neighborhood of the point ⁇ ⁇ ⁇ ⁇ ( ⁇ ).
- ⁇ p x ⁇ l / (l + (x / 0.02f) which is a decreasing function with values between 0 and 1.
- weighting term may be used as soon as they provide a low-value term in areas where the estimate provided by the term attachment to the data is unreliable.
- the weighting term may in particular include a weighting term of the homogeneous regions, for example of the form ik'g) where ⁇ is a positive increasing function on [0, + ⁇ [, A + is a function of the spatio-temporal gradients of the intensity (for example the higher eigenvalue of the gradient matrix P which is detailed below) and is a threshold value of weighting.
- ⁇ is a positive increasing function on [0, + ⁇ [, A + is a function of the spatio-temporal gradients of the intensity (for example the higher eigenvalue of the gradient matrix P which is detailed below) and is a threshold value of weighting.
- ⁇ of the form ⁇ - x ⁇ ⁇ l / (l + (0.2 / jc) 2 )
- the term ⁇ - (( v x _v x) + ( v y ⁇ v y )) of the above equation (6) is a control term.
- the regulation term makes it possible to replace the estimation of the velocity field provided by the term of attachment to the datum by the average value of said field over a neighborhood of the point ⁇ ⁇ ( ⁇ ).
- regulation term may be used, so as to provide a term for obtaining a velocity field value consistent with the estimated velocity field values in the vicinity of the point.
- the average values of the speed field can be obtained by applying a pyramidal filter F of order 5 to the speed field:
- the least squares functional J at the point ⁇ ⁇ ⁇ ⁇ ' ⁇ ⁇ ) of equation (6) is minimized by a matrix method.
- This method consists in putting the functional J in matrix form before deriving it and canceling this derivative to obtain a solution of the speed field at the iteration n. V n.
- V. corresponds to the velocity field dy y
- n calculated at the iteration corresponds to the average
- the equations defining the regularization term may be:
- R is obtained from the regularization equation (for example:
- the sums ⁇ of the matrix P can be understood as being sums on the points of the neighborhood of the point ( x o '), and the terms W and the spatio-temporal gradients of the intensity are taken at the level of said points
- a motion estimation method comprises two steps, a preliminary calculation step 100 followed by an iterative calculation step of the speed field 200.
- This series of images comprises at least one previous image I k -i and a following image I k .
- the first step 100 is to initialize the basic terms.
- It comprises a first operation 110 for calculating the spatio-temporal derivatives of the intensity from the preceding images I k _i and following I k .
- This step consists in calculating the terms in each point of the image.
- the points of the image may be chosen as corresponding to the points of the previous image, the next image or in a repository different from the two images.
- the time derivative is estimated using time interval information separating the two images.
- a fourth operation 140 makes it possible to calculate the regulation matrix R for example in the manner detailed above in equation (10).
- a fifth operation 150 may comprise the determination of a weighting level of the homogeneous regions on the basis of the eigenvalues of the matrix P.
- the homogeneous regions correspond in fact to the regions in which the eigenvalues of the matrix P are small values, and a weighting threshold value for the eigenvalues of the matrix P can therefore be fixed to allow weighting of the homogeneous regions.
- motion vectors are estimated at a plurality of points of an image selected from the preceding image I k _i and the following image Ik.
- a set 201 of values of the velocity field v y at each point of the image is firstly initialized. These values correspond to the speed field
- a second operation 220 will consist of a calculation of the calculation weighting term which can use the average values of the field determined during the first operation 210 detailed above.
- a third operation 230 consists of the resolution of the equation (10) making it possible to obtain the new velocity field + i of the step n + 1 from the matrices determined during the preliminary step 100 as well as results of the first 210 and second 220 operations of the calculation step 200.
- a fourth step 240 the opportunity to stop the iterative calculation is evaluated against a predefined criterion.
- This criterion can constitute comparing the number n of the current iteration with a maximum iteration number n max, that is to say, evaluate the condition n - n max.
- the iterative calculation step 200 ends and returns the last estimated max speed field.
- the counter of the number of iteration n is incremented during a fifth step 250 and the estimated speed field serves as initial value for a new calculation of step 200. It is therefore used as initial value for a new operation 210 for determining the averages of the speed field.
- the following operations 220, 230, 240 are then repeated.
- Another criterion for the fourth step 240 may comprise the comparison of a difference between the speed field obtained at the end of the third step 230 + i and the speed field obtained at the previous iteration.
- a cry can be to evaluate the following condition where ⁇ is a predefined value.
- the first-order development (2) of the apparent motion constraint equation (1) detailed above is valid only for small displacements, for example of the order of the pixel. To measure large displacements, it is therefore advantageous to carry out a multiscale calculation.
- the multi-scale motion estimation method 400 is applied to a series of images I ⁇ , for i integer between 1 and N, this series of images comprising at least one previous image I k -x and a following image I k .
- a first step 410 the previous images I k -x and following I k are scaled to give scaled images I k -i s , I k S ⁇
- This step may include application of an anti-aliasing filter and decimation of the image.
- An anti-aliasing filter is sometimes also referred to as an anti-aliasing filter or anti-aliasing filter and consists in removing, before decimation, frequencies greater than half the decimation frequency.
- the images may also be pre-filtered to filter the high frequencies, for example by pyramidal filters.
- one of the images for example the following image set to the scale I n s is adjusted according to the displacement field D c determined at the previous iteration, to give a next image scaled up I n s, d -
- the displacement field D c and the scale Salt are initialized to zero.
- a third step 430 comprises executing a motion estimation method according to an embodiment of the present invention comprising two steps, a preliminary calculation step 100 followed by an iterative calculation step of the speed field 200. as described above in relation to FIG. 3.
- the preceding and following images input to said method correspond to the preceding images scaling Ik-i s and following scaling recalibrated Ik S, d - This method makes it possible to estimate an estimate V of the residual velocity field between the images rescaled at a given scale.
- a fourth step 440 for converting the residual velocity field V into a residual displacement field D r is performed.
- This step may for example include a temporal integration.
- a fifth step 450 comprises the vector summation with smoothing of the residual displacement field D r and of the displacement field D c determined at the previous iteration to give a total displacement field D t , by means of the following formal equation:
- Vector summation with smoothing can include several methods of regularizing the field before or after interpolation and summation. For example, it is possible to apply a linear filter (for example a small pyramidal filter like that of equation (11)), a nonlinear filter (like a weighted average of the points of the neighborhood), a median filtering. or a non-linear diffusion with variable capacity. Variable capacity nonlinear scattering is based on a partial differential equation similar to the ches equation:
- the function c can for example be of the form c: - »l / (l + (jc / fc) 2 ) where k is a filtering capacity, for example taken equal to 1 or depending on the local noise level .
- the calculation of this equation can be done by an explicit Euler scheme with fixed pitch, the calculation of the divergence term being carried out by a Perona-Malik numerical scheme.
- a selection criterion is applied to determine whether it is necessary to re-iterate on a different scale.
- This criterion may for example be a criterion depending on the total displacement field D t obtained or a criterion depending on the number or the value of the scale Salt.
- a scaling step 470 can be executed. This step may comprise a shaping operation of the displacement field D c at the next iteration, starting from the total displacement field D t . Step 470 may further include an operation of updating the scale parameter Sel to the new value of the scale parameter.
- An alternative embodiment of the multi-scale motion estimation method 400 above may not include the second reset step 420.
- a field estimate such as that of the third step 430 is then made at each scale and the final displacement field will be composed from the fields estimated at each scale and considered reliable according to a criterion of conditioning.
- This matrix is a real symmetric matrix of size 2x2 and a criterion of good conditioning can for example be a function of the ratio between the two values
- the estimated motion field may have either a reliable value provided by a fine-scale estimate or a reliable value estimated on a coarser scale and multiplied and zoomed by the appropriate scale factor.
- an estimation method it is possible to filter the displacement field obtained temporally, the moving objects usually moving with low accelerations.
- a pyramidal filter of order 3 on recalibrated or non-recalibrated images.
- causal predictive filtering that is to say, predicting the displacement field at a given instant from the previous estimations, for example by assuming a moving speed of the moving objects in the constant image.
- Predictive causal filtering makes it possible to estimate a datum from the previous ones and a model that can possibly be adjusted.
- a classic example is Kalman filtering and the skilled person can use any known variant of the state of the art.
- G are Gabor filters, for example, of the form:
- G t (x, y) K ⁇ F- 1 [exp (- 2 ⁇ 2 ⁇ 2 ( ⁇ - f 0 cos (6>)) 2 + (yf 0 cos (6>)) 2 ⁇
- & i are the orientations of the filters and are for example chosen from 0 ° to 150 ° in steps of 30 °
- fo is the central frequency of the filters and can for example be chosen in the range of [0.25 0.4] _
- ⁇ is the range of the Gaussian and can for example be chosen equal to 1.
- FIG. 4 illustrates a system 500 for processing a series of images captured by an image sensor adapted to implement a method according to an embodiment of the present invention.
- a system 500 may advantageously correspond to an on-board electronic card.
- Such a system 500 comprises:
- a first field estimation circuit 510 may, for example, comprise a preliminary calculation circuit 511 and an iterative calculation circuit 512.
- circuits 510, 511, 512 may for example comprise microprocessors, programmable logic circuits such as FPGAs, ASICs or Application-Specifies Integrated Circuit or logic integrated circuits of another type.
- an image processing circuit 520 adapted to carry out operations of scaling or resetting of an image.
- a suitable 530 field processing circuit to perform operations on the velocity fields or displacement fields, for example the conversion of a velocity field into a motion field, the vector summation with smoothing of displacement fields or the following scaling.
- control circuit 540 adapted, on the one hand, to trigger and control a sequence of steps of a method as described above by controlling the field estimation circuits 510, image processing circuits 520 and processing of the fields 530 and, secondly, to decide whether said process is to be reiterated and to re-trigger the sequence if this is the case.
- circuits may for example comprise microprocessors, programmable logic circuits such as FPGAs, ASICs or "Application-Specific Integrated Circuit” or logic integrated circuits of another type.
- Figure 5a is indicative of a first source image with a group of four white rectangles and a small rectangle isolated to the right of the group.
- Figure 5b illustrates a second source image having the same group and single rectangle, the group of four rectangles being moved, relative to the first source image, by one pixel to the right of the image and two pixels down from the image and the isolated rectangle being moved two pixels down and two pixels to the right of the image.
- FIG. 5c represents an estimation of the horizontal velocity field obtained following the execution of a method according to an embodiment of the present invention.
- the horizontal component of the estimated motion vectors is indicated at each point of the image by a gray level intensity value varying between white (the highest value of the horizontal component of the motion vector estimated at the point of the image considered) and black (lower value of the horizontal component of the motion vector estimated at the point of the image considered).
- Fig. 5d is an illustration of an estimate of the vertical velocity field obtained following the execution of a method according to an embodiment of the present invention.
- the vertical component of the estimated motion vectors is indicated at each point of the image by a gray level intensity value varying between white (the highest value of the vertical component of the motion vector estimated at the point of the image considered) and black (lower value of the vertical component of the motion vector estimated at the point of the image considered).
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
FR1350271A FR3001073B1 (fr) | 2013-01-11 | 2013-01-11 | Estimation de mouvement d'une image |
PCT/FR2014/050049 WO2014108652A1 (fr) | 2013-01-11 | 2014-01-10 | Estimation de mouvement d'une image |
Publications (2)
Publication Number | Publication Date |
---|---|
EP2943935A1 true EP2943935A1 (fr) | 2015-11-18 |
EP2943935B1 EP2943935B1 (fr) | 2017-03-08 |
Family
ID=48692571
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
EP14703136.3A Active EP2943935B1 (fr) | 2013-01-11 | 2014-01-10 | Estimation de mouvement d'une image |
Country Status (6)
Country | Link |
---|---|
US (1) | US9510013B2 (fr) |
EP (1) | EP2943935B1 (fr) |
CA (1) | CA2897674C (fr) |
FR (1) | FR3001073B1 (fr) |
IL (1) | IL239838A (fr) |
WO (1) | WO2014108652A1 (fr) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10499870B2 (en) * | 2017-05-19 | 2019-12-10 | The Chinese University Of Hong Kong | Methods and apparatuses for quantifying vascular fluid motions from DSA |
US10564174B2 (en) * | 2017-09-06 | 2020-02-18 | Pixart Imaging Inc. | Optical sensing apparatuses, method, and optical detecting module capable of estimating multi-degree-of-freedom motion |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2648254B2 (fr) * | 1988-09-23 | 1991-08-30 | Thomson Csf | Procede et dispositif d'estimation de mouvement dans une sequence d'images animees |
US8457410B2 (en) * | 2007-02-14 | 2013-06-04 | Technion Research And Development Foundation Ltd. | Over-parameterized variational optical flow method |
US9307251B2 (en) * | 2009-08-19 | 2016-04-05 | Sharp Laboratories Of America, Inc. | Methods and systems for determining data-adaptive weights for motion estimation in a video sequence |
-
2013
- 2013-01-11 FR FR1350271A patent/FR3001073B1/fr not_active Expired - Fee Related
-
2014
- 2014-01-10 EP EP14703136.3A patent/EP2943935B1/fr active Active
- 2014-01-10 US US14/760,414 patent/US9510013B2/en active Active
- 2014-01-10 CA CA2897674A patent/CA2897674C/fr active Active
- 2014-01-10 WO PCT/FR2014/050049 patent/WO2014108652A1/fr active Application Filing
-
2015
- 2015-07-08 IL IL239838A patent/IL239838A/en active IP Right Grant
Also Published As
Publication number | Publication date |
---|---|
US9510013B2 (en) | 2016-11-29 |
IL239838A0 (en) | 2015-08-31 |
FR3001073B1 (fr) | 2016-05-06 |
FR3001073A1 (fr) | 2014-07-18 |
IL239838A (en) | 2016-11-30 |
WO2014108652A1 (fr) | 2014-07-17 |
EP2943935B1 (fr) | 2017-03-08 |
CA2897674A1 (fr) | 2014-07-17 |
CA2897674C (fr) | 2016-07-05 |
US20150358637A1 (en) | 2015-12-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP3138079B1 (fr) | Procédé de suivi de forme dans une scène observée par un capteur asynchrone de lumière | |
EP3221841B1 (fr) | Procede et dispositif de filtrage adaptatif temps reel d'images de disparite ou de profondeur bruitees | |
EP3213033A1 (fr) | Procédé d'estimation d'un état de navigation contraint en observabilité | |
FR2882160A1 (fr) | Procede de capture d'images comprenant une mesure de mouvements locaux | |
EP3272119A1 (fr) | Procede de reconstruction 3d d'une scene | |
EP1746486B1 (fr) | Procédé de détection de déplacement d'une entité pourvue d'un capteur d'images et dispositif pour le mettre en oeuvre | |
EP1998288A1 (fr) | Procédé de détermination du déplacement d'une entité pourvue d'un capteur de séquence d'images, programme d'ordinateur, module et souris optique associés | |
EP3301644B1 (fr) | Procédé de construction d'une carte de profondeur d'une scène et/ou d'une image entièrement focalisée | |
EP2943935B1 (fr) | Estimation de mouvement d'une image | |
CA3105372C (fr) | Traitement d'un bruit impulsionnel dans une sequence video | |
EP3701492B1 (fr) | Procede de restauration d'images | |
EP3072110B1 (fr) | Procédé d'estimation du mouvement d'un objet | |
FR2955007A1 (fr) | Estimation de mouvement global et dense | |
EP4341897B1 (fr) | Procédé et dispositif de traitement d'une séquence d'images pour la détermination de vignettes poursuivies dans ladite séquence d'images | |
FR3066633A1 (fr) | Procede de defloutage d'une image | |
EP4386673A1 (fr) | Méthode d'apprentissage non supervisé d'un modèle d'estimation de flux optique | |
WO2022233977A1 (fr) | Procédé de traitement d'images | |
FR3051066A1 (fr) | Procede de restauration d'images | |
WO2019115324A1 (fr) | Procede de synthese d'images | |
FR3146534A1 (fr) | Procede d’estimation d’une carte de profondeur d’une image monoculaire et dispositifs associes | |
FR3103940A1 (fr) | Procédé et dispositif de traitement d’images | |
Hutchison | Signed coded exposure sequences for velocity and shape estimation from a single photo | |
WO2015052290A1 (fr) | Estimation statistique de paramètres d'égalisation d'image numérique | |
FR2993690A1 (fr) | Procede de determination d'un vecteur de mouvement ou d'un modele de mouvement entre deux images numeriques |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
17P | Request for examination filed |
Effective date: 20150702 |
|
AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
AX | Request for extension of the european patent |
Extension state: BA ME |
|
DAX | Request for extension of the european patent (deleted) | ||
RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: SAFRAN ELECTRONICS & DEFENSE |
|
RIC1 | Information provided on ipc code assigned before grant |
Ipc: H04N 19/43 20140101ALI20160719BHEP Ipc: G06T 7/20 20060101AFI20160719BHEP Ipc: H04N 5/14 20060101ALI20160719BHEP Ipc: H04N 19/513 20140101ALI20160719BHEP |
|
GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
INTG | Intention to grant announced |
Effective date: 20160916 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: GRANT OF PATENT IS INTENDED |
|
GRAS | Grant fee paid |
Free format text: ORIGINAL CODE: EPIDOSNIGR3 |
|
GRAA | (expected) grant |
Free format text: ORIGINAL CODE: 0009210 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE PATENT HAS BEEN GRANTED |
|
AK | Designated contracting states |
Kind code of ref document: B1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
REG | Reference to a national code |
Ref country code: GB Ref legal event code: FG4D Free format text: NOT ENGLISH |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: EP Ref country code: AT Ref legal event code: REF Ref document number: 874144 Country of ref document: AT Kind code of ref document: T Effective date: 20170315 |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: FG4D Free format text: LANGUAGE OF EP DOCUMENT: FRENCH |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R096 Ref document number: 602014007386 Country of ref document: DE |
|
REG | Reference to a national code |
Ref country code: NL Ref legal event code: FP |
|
REG | Reference to a national code |
Ref country code: SE Ref legal event code: TRGR |
|
REG | Reference to a national code |
Ref country code: LT Ref legal event code: MG4D |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: FI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: GR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170609 Ref country code: NO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170608 Ref country code: HR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: LT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
REG | Reference to a national code |
Ref country code: AT Ref legal event code: MK05 Ref document number: 874144 Country of ref document: AT Kind code of ref document: T Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: RS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: LV Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: BG Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170608 Ref country code: ES Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: AT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: EE Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: SK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: CZ Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: RO Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SM Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: IS Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170708 Ref country code: PT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170710 Ref country code: PL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
REG | Reference to a national code |
Ref country code: DE Ref legal event code: R097 Ref document number: 602014007386 Country of ref document: DE |
|
REG | Reference to a national code |
Ref country code: FR Ref legal event code: PLFP Year of fee payment: 5 |
|
PLBE | No opposition filed within time limit |
Free format text: ORIGINAL CODE: 0009261 |
|
STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: DK Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
26N | No opposition filed |
Effective date: 20171211 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: SI Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
REG | Reference to a national code |
Ref country code: CH Ref legal event code: PL |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MT Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LU Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180110 |
|
REG | Reference to a national code |
Ref country code: IE Ref legal event code: MM4A |
|
REG | Reference to a national code |
Ref country code: BE Ref legal event code: MM Effective date: 20180131 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: LI Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180131 Ref country code: CH Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180131 Ref country code: BE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180131 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: IE Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20180110 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: MC Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: TR Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: CY Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 Ref country code: HU Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT; INVALID AB INITIO Effective date: 20140110 Ref country code: MK Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES Effective date: 20170308 |
|
PG25 | Lapsed in a contracting state [announced via postgrant information from national office to epo] |
Ref country code: AL Free format text: LAPSE BECAUSE OF FAILURE TO SUBMIT A TRANSLATION OF THE DESCRIPTION OR TO PAY THE FEE WITHIN THE PRESCRIBED TIME-LIMIT Effective date: 20170308 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: GB Payment date: 20231219 Year of fee payment: 11 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: SE Payment date: 20231219 Year of fee payment: 11 Ref country code: NL Payment date: 20231219 Year of fee payment: 11 Ref country code: FR Payment date: 20231219 Year of fee payment: 11 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: DE Payment date: 20231219 Year of fee payment: 11 |
|
PGFP | Annual fee paid to national office [announced via postgrant information from national office to epo] |
Ref country code: IT Payment date: 20240102 Year of fee payment: 11 |